Computational Design of Strip Waveguide Bragg Gratings Using Neural Networks for Hexagonal Shaped Refractive Index Biosensors Based Blood Sample Detection

Biosensors using Waveguide Bragg Grating (WBG) technology have gained a lot of attention due to their many useful properties, including their sensitivity, robustness, and suitability for lab-on-a-chip applications. These biosensors must be designed from Maxwell’s equations for optical waveguide geometries and numerical solutions for these involve formidable calculations; optimizing is therefore not easy or fast. To overcome these challenges, a hexagon-shaped refractive index-based biosensing approach is introduced for strip WBG structures, leveraging Bayesian Maximum-Entropy Asymmetric Regularized Quantized Decision Osprey Optimization Networks. This work combines Bayesian asymmetric quantized neural networks (BAQNN) with a Maximum-Entropy Regularized Decision Transformer (MERDT) alongside with the use of Osprey Optimization Algorithm (OOA) to enhance the accuracy of the sensing parameters estimate. The method achieves over 99.9% mode classification accuracy and predicts effective refractive index ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:msub> <mml:mrow> <mml:mi>m</mml:mi> </mml:mrow> <mml:mrow> <mml:mi mathvariant="italic">eff</mml:mi> </mml:mrow> </mml:msub> </mml:math> ) having a 0.5% mean absolute error (MAE), while sensitivity, quality factor, and reflectivity predictions exhibit MAEs below 3% with a confinement loss of 0.06. Applying this model to infection detection, this approach leads to faster design cycles than required for sensor development, decreases computation time, and allows for effective optimization, making it valuable for future employment in various biosensor design.